Share your thoughts, 1 month free Claude Pro on usSee more

STEM / Multimodal Reasoning on MathVista (Accuracy)

83.97Accuracy

STEP3-VL-10B

Updated 1mo ago

Evaluation Results

Method	Links
STEP3-VL-10B 2026.01		83.97
GLM-4.6V Flash 2026.01		82.85
MiMo-VL RL-2508 2026.01		79.86
Qwen3-VL Thinking 2026.01		78.5
InternVL 3.5 2026.01		76.78
CodePercept-32B-S1 2026.03		75.9
Gemini2.5-Pro 2026.03		74.8
Qwen3-VL-32B-Instruct 2026.03		74
Qwen3-VL-235A22B-Instruct 2026.03		73.8
CodePercept-32B-S1 2026.03		72.9
Qwen3-VL-32B-Instruct 2026.03		72.2
CodePercept-8B-S1 2026.03		71.4
CodePercept-4B-S1 2026.03		71.3
Claude-Opus 4.1-Thinking 2026.03		71.1
Qwen3-VL-8B-Instruct 2026.03		71
CodePercept-8B-S1 2026.03		70.2
CodePercept-4B-S1 2026.03		69.6
Qwen3-VL-8B-Instruct 2026.03		69.6
Qwen3-VL-4B-Instruct 2026.03		69.2
Qwen3-VL-30A3B-Instruct 2026.03		68
InternVL3.5-8B 2026.03		67.7
Qwen2.5-VL-72B 2026.03		67.5
Qwen3-VL-4B-Instruct 2026.03		67.3
MiniCPM-V-4.5 2026.03		66.6
Intern-S1-8B 2026.03		65.7
GPT5-Thinking 2026.03		65.2
KeyeVL1.5-8B 2026.03		64.9
GLM-4.1V-9B 2026.03		64.6